Dokument: Cerebellar Contributions to Prediction Error Processing in Reinforcement Learning
| Titel: | Cerebellar Contributions to Prediction Error Processing in Reinforcement Learning | |||||||
| URL für Lesezeichen: | https://docserv.uni-duesseldorf.de/servlets/DocumentServlet?id=70943 | |||||||
| URN (NBN): | urn:nbn:de:hbz:061-20251017-131436-5 | |||||||
| Kollektion: | Dissertationen | |||||||
| Sprache: | Englisch | |||||||
| Dokumententyp: | Wissenschaftliche Abschlussarbeiten » Dissertation | |||||||
| Medientyp: | Text | |||||||
| Autor: | Huvermann, Dana Maria [Autor] | |||||||
| Dateien: |
| |||||||
| Beitragende: | Prof. Dr. Peterburs, Jutta [Gutachter] Prof. Dr. Bellebaum, Christian [Gutachter] | |||||||
| Dewey Dezimal-Klassifikation: | 100 Philosophie und Psychologie » 150 Psychologie | |||||||
| Beschreibungen: | Reinforcement learning is an essential function for most higher organisms to survive and thrive in dynamic environments. Herein, reward prediction errors (RPEs) play a central role. RPEs signal if an action outcome deviated from expectations and help improve future predictions. Traditionally, RPE processing has been linked to midbrain and prefrontal regions, but more recent evidence indicates RPE processing in the cerebellum as well. This dissertation explored the role of the cerebellum in reinforcement learning using different models of cerebellar dysfunction (i.e., patients with cerebellar stroke, transcranial magnetic stimulation of the cerebellum in healthy individuals) while recording cerebral signals. The key question was whether cerebellar deficits would change cerebral processing of RPEs.
Manuscript 1 investigated whether cerebellar deficits in patients with chronic cerebellar strokes and in healthy young adults receiving single-pulse cerebellar transcranial magnetic stimulation (TMS) would result in deficient RPE processing in a reinforcement learning task. Electroencephalography (EEG) was employed to measure cerebral RPE processing, focussing on the feedback-related negativity (FRN). The FRN had previously been shown to not only distinguish between negative and positive feedback but to also covary with RPEs. Both experiments showed that a cerebellar dysfunction led to a blunted RPE reflection in the FRN, suggesting cerebellar contributions to cerebral RPE processing. No behavioural deficits in overall learning success were apparent, potentially indicating compensation by other brain areas. Manuscript 2 investigated response error processing in reinforcement learning. While naïve learners rely on feedback, advanced learners are able to recognise errors already at the response execution. Previous studies performed in non-reinforcement learning contexts implied the cerebellum in error processing. Manuscript 2 investigated whether cerebellar TMS applied in healthy adults altered error processing in reinforcement learning. EEG analyses of the error-related negativity (ERN) and error positivity (Pe) in the response-locked event-related potential showed that cerebellar TMS blunted error processing in the ERN but enhanced it in the Pe, with the latter potentially indicating compensation. Last, Manuscript 3 explored the timing aspect of cerebellar-cerebral communication in error processing. While for the motor domain, a clear time window for cerebellar-cerebral communication is known, this is less clear for the cognitive domain. Several TMS timings around individual ERN latencies were investigated in a Go/Nogo Flanker task, i.e., a response conflict and inhibition task. It could be shown that only stimulation near ERN latency was effective in reducing error processing in ERN. Together, these findings further implicate the cerebellum in reinforcement learning and the more general area of performance monitoring. Cerebellar deficits altered cerebral processing, even though this did not directly translate to substantial behavioural impairments. Future research into the functional significance of these cerebellar deficits is required.Reinforcement learning is an essential function for most higher organisms to survive and thrive in dynamic environments. Herein, reward prediction errors (RPEs) play a central role. RPEs signal if an action outcome deviated from expectations and help improve future predictions. Traditionally, RPE processing has been linked to midbrain and prefrontal regions, but more recent evidence indicates RPE processing in the cerebellum as well. This dissertation explored the role of the cerebellum in reinforcement learning using different models of cerebellar dysfunction (i.e., patients with cerebellar stroke, transcranial magnetic stimulation of the cerebellum in healthy individuals) while recording cerebral signals. The key question was whether cerebellar deficits would change cerebral processing of RPEs. Manuscript 1 investigated whether cerebellar deficits in patients with chronic cerebellar strokes and in healthy young adults receiving single-pulse cerebellar transcranial magnetic stimulation (TMS) would result in deficient RPE processing in a reinforcement learning task. Electroencephalography (EEG) was employed to measure cerebral RPE processing, focussing on the feedback-related negativity (FRN). The FRN had previously been shown to not only distinguish between negative and positive feedback but to also covary with RPEs. Both experiments showed that a cerebellar dysfunction led to a blunted RPE reflection in the FRN, suggesting cerebellar contributions to cerebral RPE processing. No behavioural deficits in overall learning success were apparent, potentially indicating compensation by other brain areas. Manuscript 2 investigated response error processing in reinforcement learning. While naïve learners rely on feedback, advanced learners are able to recognise errors already at the response execution. Previous studies performed in non-reinforcement learning contexts implied the cerebellum in error processing. Manuscript 2 investigated whether cerebellar TMS applied in healthy adults altered error processing in reinforcement learning. EEG analyses of the error-related negativity (ERN) and error positivity (Pe) in the response-locked event-related potential showed that cerebellar TMS blunted error processing in the ERN but enhanced it in the Pe, with the latter potentially indicating compensation. Last, Manuscript 3 explored the timing aspect of cerebellar-cerebral communication in error processing. While for the motor domain, a clear time window for cerebellar-cerebral communication is known, this is less clear for the cognitive domain. Several TMS timings around individual ERN latencies were investigated in a Go/Nogo Flanker task, i.e., a response conflict and inhibition task. It could be shown that only stimulation near ERN latency was effective in reducing error processing in ERN. Together, these findings further implicate the cerebellum in reinforcement learning and the more general area of performance monitoring. Cerebellar deficits altered cerebral processing, even though this did not directly translate to substantial behavioural impairments. Future research into the functional significance of these cerebellar deficits is required. | |||||||
| Lizenz: |  Dieses Werk ist lizenziert unter einer Creative Commons Namensnennung 4.0 International Lizenz | |||||||
| Fachbereich / Einrichtung: | Mathematisch- Naturwissenschaftliche Fakultät » WE Psychologie | |||||||
| Dokument erstellt am: | 17.10.2025 | |||||||
| Dateien geändert am: | 17.10.2025 | |||||||
| Promotionsantrag am: | 13.05.2025 | |||||||
| Datum der Promotion: | 01.10.2025 |
